Update: Okay as you explained more, now here is some code works for the sample data you give. For me, the best way to interpret your requirement is, to base everything on the sample you gave, so you may have to make some changes, to make it fit:

use strict;

open(IN, "<", "data.txt");
my @in = <IN>;
close(IN);

open(OUT, ">", "out.txt");
my $out = "";   
foreach my $in (@in) {
    chomp($in);
    if ($in =~ m/^(.*?)\\(.*?)\s*$/) {
        $out .= " $2";
    } else {
        if ($out ne "") {
            print OUT "$out\n";
        }
        $in =~ m/^(.*?)\s*$/;
        $out = $1;
    }
}
print OUT "$out\n";
close(OUT);
[download]

Oringinal Reply: The way you store your data is really problematic. Use a flatened structure like _section1, _section2, _section12, _section13 etc is not a good idea at all.

It is much more natural to look at your data structure as a tree. For this purpose, there are lots of implementation choices.

One choice is array-ref of array-ref... So first you would have, $source{_source} which is an array ref. For example, your _section1, would be

$source{_source}->[0]
[download]

If your case is that you need to store some text for section1 itself, even though it is not a leaf, then store those things in:

$source{_source}->[0][0]#dedicate 0 to myself
[download]

If this section1 is not a leaf, itself would be another array-ref, and its first child would be

$source{_source}->[0][1]#my 1st child
[download]

This goes on and on.

There is a big chance that you could make your parse function recursive.

A piece of sample code:

use Data::Dumper;

use strict;

my %source;
$source{_source} = [];
$source{_source}->[0] = [];
$source{_source}->[0][0] = "section1";#myself
$source{_source}->[0][1] = "section11";#1st child
$source{_source}->[0][1] = "section12";#2nd child

print Dumper(\%source);
[download]